Recently I got two sleepless nights, which is terrible for me, but good for my program. The result was a late move reduction (at most ome on the search path) that finally seems to work.
The question is: suppose that You get a move eligible for LMR exactly one ply before your futility threshold. Would You like to retain the option of testing such a move for futility immediately after it becomes a candidate for a reduction?
regards,
pawel koziol
LMR and futility
Moderators: hgm, Rebel, chrisw
Re: LMR and futility
I think that futility and lmr are not exclusive, both heuristics want to prune fail low branches ; futility checks a 'real' score, and lmr checks some magic statistics.
But your hash table can contain some important information on the position, so it's surely better to play the move normaly with the reduced depth rather than immedialty doing your futility check.
HJ.
But your hash table can contain some important information on the position, so it's surely better to play the move normaly with the reduced depth rather than immedialty doing your futility check.
HJ.